The main risks are factual inaccuracies based on thin or misread notes, tone mismatches that make the proposal feel generic or off-brand, and emphasis errors that lead with the wrong benefits for this particular prospect — all of which can damage a relationship that took real effort to build.
What Can Actually Go Wrong
An AI agent working from incomplete notes will fill gaps with inference. If your notes say “scaling challenge” without specifying what that means, the agent might infer a revenue-scaling problem when the prospect actually meant team-scaling. The proposal then addresses the wrong problem entirely. The prospect reads it and thinks you weren’t listening — which is the opposite of what you want.
Tone errors are subtler but equally damaging. An agent tuned toward professional formality might produce a proposal that reads like a corporate services agreement when your prospect expects the warm, direct communication you brought to the call. That mismatch creates cognitive dissonance — the proposal and the person they spoke to feel like different entities. Trust erodes.
The third category is emphasis error. Your prospect said three things. One of them was the real concern; two were surface-level descriptions. A human who was on the call knows which one was the real concern. An agent reading notes doesn’t — and if it leads with the wrong one, the proposal feels off even if everything is technically accurate.
The Review Requirement Is Not Optional
Every credible use of AI in client-facing work includes a human review step. This isn’t about distrust of AI — it’s about understanding what AI is good at (producing structured, well-written drafts quickly) and what it isn’t (knowing this specific person and relationship the way you do). The agent generates; you verify and approve.
A five-minute review eliminates the vast majority of the risk. The specific things to check: does every factual claim in the proposal accurately reflect your call notes? Does the tone match your relationship with this prospect? Is the primary emphasis on the thing this person actually cares most about? Those three checks catch 90% of the errors an unreviewed proposal would contain.
What This Means for Educators
Coaches and consultants who work on high-trust, high-value engagements have the most to lose from an unreviewed AI error. A $10,000 consulting engagement or a year-long coaching relationship is worth far more than the five minutes it takes to review the draft. Build the review step in and keep it non-negotiable.
The Simple Rule
AI generates; you approve. No client-facing output from an agent goes out without your eyes on it. That discipline keeps the efficiency of automation and the quality of human judgment working together, rather than trading one for the other.